可以預測網頁連結的智慧型搜尋引擎之設計

An Intelligent Search Engine for Hyperlink Prediction

陳林志(Cayley Chen) ； 陸承志(Cheng-Jye Luh)

《資訊管理學報》 12卷1期 (2005/01) Pp. 111-130

https://doi.org/10.6382/JIM.200501.0111

摘要

本文提出一個依據使用者行為進行預測網頁的搜尋引擎，它具有兩種搜尋機制：搜尋引擎投票向量(SVV)以及連結預測(HLP)。SVV的方法是對六個知名的搜尋引擎進行英文關鍵字搜尋，當一個網頁能夠從這些搜尋引擎取得較多且較前面的排名，則該網頁的權重自然較高。至於HLP的方法是根據SVV的結果進行深度連結預測，以推薦使用者最可能點擊的連結。經過初步的評估，使用者對於我們搜尋引擎的滿意度高於其他一般搜尋引擎。本搜尋引擎目前只支援英文關鍵字的搜尋，未來將增加中文關鍵字的搜尋功能。

關鍵字

向量投票 ； 連結預測 ； 搜尋引擎 ； 使用者行為函數

並列摘要

This paper presents an intelligent search engine which is capable of predicting web pages on the base of user preferences. Two search methods are implemented in this search engine. First, a search engine vector voting (SVV) method is developed to rank the web pages returned from six well known search engines. A web page will be ranked high if it is listed near the top in several search engine results. Based on the URLs in SVV results, a hyperlink prediction (HLP) method is then developed for predicting all the hyperlinks on which users are most likely to click. Preliminary user study results indicate that users are more satisfied with our search engine than others. The work presented in this paper is mainly on English keyword search, and Chinese keyword search will be developed in future research.

並列關鍵字

Search Engine Vector Voting ； Hyperlink Prediction ； MetaSearch ； User Behavior Function

參考文獻

Chakrabarti, S.(1999).Mining the Web`s Link Structure.IEEE Internet Computing.60-67.

Dreilinger, D.,A. Howe(1996).An Information Gathering Agent for Querying Web Search Engine.Tech. Report CS-96-l11, Computer Science Dept., Colorado state Univ., Fort Collins, Cob..(Tech. Report CS-96-l11, Computer Science Dept., Colorado state Univ., Fort Collins, Cob.).,未出版.

Goldberg, D. E.(1999).(Genetic Algorithms in Search, Optimization and Machine Learning).

Harman, D.(1992).Ranking Algorithm.(Information Retrieval Data Structure and Algorithm).

Henzinger, M. R.(2001).Hyperlink Analysis for the Web.IEEE Internet Computing.45-50.